【作者】 高敏杰；

【导师】 詹晓北； 史仲平；

【作者基本信息】 江南大学 ， 发酵工程， 2012， 博士

【摘要】 甲醇营养型毕赤酵母(Methylotrophic Pichia pastoris)是近年来应用广泛、发展迅速的一种外源蛋白表达系统。实现目标外源蛋白(猪α干扰素,pIFN-α;人血清白蛋白-人白介素-2融合蛋白,IL-2-HSA)的高效表达离不开有效的过程控制策略。然而,发酵过程的控制和优化一般建立在非构造式动力学模型的基础上,难以适应发酵过程的时变特征,严重制约了过程控制和最优化系统的有效性和通用能力。为进一步改善毕赤酵母培养过程的控制性能,首先要深入理解毕赤酵母细胞在不同阶段的代谢特征和生理状态。然后结合智能模式识别控制和宏观的代谢调控等方法,针对毕赤酵母发酵过程的不同阶段,建立有效、通用的在线控制和故障诊断方法。基于以上的学术思想,本论文主要开展了以下研究:(1)在5 L发酵罐中研究了不同的甘油流加策略对毕赤酵母细胞生长和蛋白(pIFN-α)表达的影响。传统的DO-Stat流加策略导致细胞的比生长速率很低,在后续的诱导阶段细胞干重只有90 g/L。采用基于DO/pH在线测量和人工神经网络模式识别模型的控制系统(Artificial Neural Network Pattern Recognition based Control, ANNPR-Ctrl),可以把细胞浓度提高到120 g-DCW/L,pIFN-α的浓度也从采用DO-Stat流加策略时的0.43 g/L提高到了0.95 g/L。但此方法在甘油流加阶段的最后10 h,细胞的比生长速率由0.15~0.18 h-1降至0.03~0.08 h-1,导致pIFN-α产量的不稳定。该问题通过改进型的ANNPR-Ctrl系统(Imp-ANNPR-Ctrl)得以解决。新的流加策略使细胞的比生长速率维持在0.11 h?1以上,相应的pIFN-α的浓度达到了1.43 g/L的水平,是采用标准的ANNPR-Ctrl控制系统时pIFN-α浓度的1.5倍。研究结果表明,提高培养期细胞的比生长速率有利于诱导期外源蛋白的表达。通过Imp-ANNPR-Ctrl流加策略,细胞浓度和比生长速率都得到了提高,pIFN-α的生产性能也相应改善。(2)前期研究对比了不同甲醇浓度水平(5 g/L, 10 g/L和20 g/L)对pIFN-α表达的影响。结果表明甲醇浓度10 g/L条件下,pIFN-α的表达水平最高为1.25 g/L。而在此条件下,诱导初期OUR/DO的变化模式显示细胞有6 h左右的适应期,说明这一阶段甲醇浓度过高。为了缩短细胞适应期,提高蛋白表达水平,我们提出了两阶段甲醇浓度控制策略。即甲醇浓度在诱导前20 h从0 g/L线性增加到10 g/L,在20 h后维持在10 g/L的水平。在此条件下,pIFN-α的最高浓度达到了1.81 g/L。(3)在毕赤酵母表达外源蛋白过程中,诱导期甲醇浓度是最重要的控制参数,如何准确有效的把甲醇浓度控制在合适的范围内直接决定了蛋白表达量的高低。在研究氧气消耗速率(OUR)和pIFN-α表达水平以及甲醇浓度的相互关系的基础上,我们发现,正常发酵时,诱导期OUR稳定在200~250 mmol/L/h的水平。当甲醇浓度突然升高时(>20 g/L),OUR迅速下降,出现转折点。通过在线监控OUR的变化模式,我们建立了一种简单的在线故障诊断方法。该方法能够在线及时准确地识别甲醇浓度过高的问题,pIFN-α表达的稳定性得到进一步增强。(4)单一过程参数(OUR)进行故障诊断不具有通用性。通过选取多个在线参数(OUR,CER,搅拌转速,甲醇流加量,氨水流加量等),我们提出了基于自联想人工神经网络的毕赤酵母表达IL-2-HSA过程的甲醇诱导期两阶段故障诊断方法。该方法能够在线及时准确地识别甲醇电极出现的各种故障和pH值的漂移。当系统提示出现故障时,离线分析,并对比最优化的pH值和甲醇浓度变化曲线,确定故障类型,采取相应措施。当出现甲醇浓度过高的故障时,通过以2 g/L/h的速率限制性添加甘油可以有效提高菌体的呼吸活性,促进菌体生长,缓解甲醇过量积累对细胞的毒害作用。(5)在前期5 L发酵罐研究的基础上,我们开展了10 L发酵罐规模的放大研究。结果表明,采用相同的甘油流加和甲醇诱导控制策略,pIFN-α在10 L发酵罐中的表达水平明显低于5 L罐。其主要原因是蛋白表达的时间缩短,降解严重,细胞代谢活性较低。为解决上述问题,我们采用了低温(20℃)诱导策略。在此条件下,毕赤酵母的最大耐甲醇能力达到约40 g/L,有效地解决了因甲醇浓度变化造成的系统不稳定问题。此外,通过甲醇代谢和能量代谢分析计算表明,在20℃下,更多的碳流用于pIFN-α蛋白的合成;ATP再生效率提高了49~66%;pIFN-α最高浓度水平从0.29 g/L提高到1.1 g/L (比活性1.4×10~6 IU/mg)。10 L发酵罐中pIFN-α表达水平有了明显提高,基本达到5 L罐的水平。(6)降低诱导温度能够有效提高毕赤酵母发酵生产pIFN-α的性能。但低温条件下,大量冷却水的消耗和氧气消耗速率的提高导致发酵成本明显上升。在10 L发酵罐和不同诱导策略(30℃/20℃甲醇单独诱导和30℃甲醇/山梨醇混合流加)条件下,我们开展了过程参数和代谢酶学分析以及能量代谢过程解析。结果表明:当采用30℃甲醇/山梨醇混合流加时,pIFN-α合成的主要供能途径由甲醇诱导时的甲醛异化产能途径转向混合流加时的TCA循环。混合流加条件下,甲醛异化代谢途径被弱化,有毒代谢中间产物甲醛的积累得到有效缓解;理论耗氧速率大幅下降,整个诱导阶段溶氧浓度处在适中的水平;能量和甲醇的利用效率显著提高,更多的甲醇可以直接用于细胞和蛋白的合成。此时,pIFN-α浓度达到2.1 g/L,比活性8.6×10~6 IU/mg。另外,甲醇/山梨醇混合流加可以在30℃常温和使用空气供氧的条件下进行,发酵操作成本明显降低,发酵性能得到提高。更多还原

【Abstract】 The methylotrophic yeast Pichia pastoris is currently one of the most effective and versatile systems for the expression of heterologous proteins. Normally, the fermentation process control strategry is based on unstructured models and difficult to apply to dynamic changing of fermentation which largely prevents its effectiveness and versatility. In this dissertation, to improve the fermentation performance of Pichia pastoris, by combinational using techniques of artificial intelligent and metabolic/enzymatic analysis, new control methods were established for the different Pichia pastoris fermentation process stages. The main results of this dissertation were summarized as following:(1) Effective expression of porcine interferon-α(pIFN-α) with recombinant Pichia pastoris was conducted in a 5 L fermentor. The influence of the glycerol feeding strategy on the specific growth rate and protein production was investigated. The traditional DO-Stat feeding strategy led to very low cell growth rate resulting in low dry cell weight (DCW) of about 90 g/L during the subsequent induction phase. The previously reported Artificial Neural Network Pattern Recognition (ANNPR) model based glycerol feeding strategy improved the cell density to 120 g-DCW/L. pIFN-αconcentration improved to 0.95 g/L from only 0.43 g/L with DO-Stat feeding strategy. With ANNPR-Ctrl model, the specific growth rate decreased from 0.15-0.18 h-1 to 0.03～0.08 h-1 during the last 10 h of the glycerol feeding stage leading to a variation of the pIFN-αproduction as the glycerol feeding scheme had a significant effect on the induction phase. This problem was resolved by an improved ANNPR model based feeding strategy to maintain the specific growth rate above 0.11 h-1. With this feeding strategy, the pIFN-αconcentration reached 1.43 g/L, about 1.5 folds of that obtained with the previously adopted feeding strategy. Our results showed that, increasing the specific growth rate favored the target protein production and the glycerol feeding methods directly influenced the induction stage. Consequently, higher cell density and specific growth rate as well as effective pIFN-αproduction have been achieved by our novel glycerol feeding strategy.(2)Our previous study compared the results of setting constant methanol concentration of different levels (5 g/L, 10 g/L and 20 g/L) during the induction phase and concluded that the highest pIFN-αactivity was achieved at methanol concentration of 10 g/L. The changing patterns of DO/OUR indicated an adaptation phase of 6 h, which showed the methanol concentration of 10 g/L was too high for the beginning of the induction. Therefore, effective expression of pIFN-αwith recombinant Pichia pastoris was conducted in the 5 L bench-scale fermentor using an on-line methanol electrode based feeding process with the control level of methanol concentration linearly increased to 10 g/L for the first 20 h and maintained at 10 g/L for the rest expression phase. With this two-stage control process, the highest pIFN-αconcentration reached a level of 1.81 g/L.(3) For the protein expression with Pichia pastoris, methanol concentration during induction phase was the most important parameter dominating heterologous protein production and should be strictly controlled at adequate levels. The relationship between OUR methanol concentration and pIFN-αactivity was analysed. Normally, the OUR level stayed at 200～250 mmol/L/h. The OUR changing patterns showed a turning point when methanol concentration increased largely. Consequently, the pIFN-αexpression stability could be further enhanced with the aid of a simple on-line fault diagnosis method for methanol over-feeding based on oxygen uptake rate (OUR) changing patterns.(4) Fault diagnosis only with OUR can not be used widely. In this chapter, based on the effective recognition of physiological status and characteristics of parameters, an autoassociative neural network (AANN) model was used for two-stage fault diagnosis in the process of IL-2-HSA expression with Pichia pastoris. The optimized AANN could provide on-line and accurate fault alarm for Pichia pastoris induction stage. It was potentially helpful in supplying useful information for removing fault and recovering abnormal fermentation. When detecting methanol over-feeding, glycerol limited feeding at 2 g/L/h could improve the cell activity and release the toxicity of methanol.(5) The scaled-up pIFN-αproduction was conducted in a 10 L fermentor. Because of short expression phase, protein degradation and low cell activity, the pIFN-αactivities obtained in 10 L fermentor were obviously lower than those in the 5 L fermentor with the same control strategies. A low temperature induction strategy at 20oC was thus adopted for efficient pIFN-αproduction in a 10 L fermentor. With the strategy, maximal methanol tolerance level could reach about 40 g/L to effectively deal with methanol concentration variations, so that the complicated on-line methanol measurement system could be eliminated. Moreover, metabolic analysis based on multiple state-variables measurements indicated that pIFN-αantiviral activity enhancement profited from the formation of an efficient ATP regeneration system at 20oC induction and the enhanced carbon flow towards to pIFN-αsynthesis. Compared to the induction strategy at 30oC, the proposed strategy increased the ATP regeneration rate by 49-66%, the maximal pIFN-αconcentration reached 1.1 g/L and the specific antiviral activity was 1.4×10~6 IU/mg.(6) It has been reported that, an enhanced pIFN-αfermentative production by Pichia pastoris could be achieved when inducing at a lower temperature of 20oC. However, induction at low temperature leads to a high operation cost including a large amount of cooling water and pure oxygen usage. In this study, an alternative pIFN-αproduction mode using sorbitol/methanol co-feeding strategy at room temperature of 30oC was conducted in a 10 L fermentor with the focus on analyzing changing patterns of energy regeneration and the corresponding metabolic enzymology, aiming at improving fermentation performance and reducing operation cost simultaneously. The results showed that when using the methanol/sorbitol co-feeding strategy at 30°C, major energy metabolism energizing pIFN-αsynthesis shifted from formaldehyde dissimilatory energy metabolism pathway to TCA cycle. Under this operation mode, the formaldehyde dissimilatory pathway was weakened and accumulation of toxic intermediate metabolite-formaldehyde was relieved; the theoretical oxygen consumption rate was largely reduced, leading to a moderate DO level throughout the induction phase; energy/methanol utilization efficiency was largely increased so that more methanol could be directed into the cell/protein synthesis route. As a result, pIFN-αconcentration reached the highest level of 2.1 g/L which was about 1.9～7.2 folds of those obtained under pure methanol induction at 20°C and 30°C, respectively. More importantly, enhanced pIFN-αproduction by sorbitol/methanol co-feeding strategy could be implemented under mild operation conditions at room temperature and using air for aeration, which greatly reduced fermentation costs and improved the entire fermentation performance in turn.更多还原